I don’t usually talk about this book. The title has been out of print for eight years. Most of our current authors and readers have never heard of it. But I’ve been thinking about it lately, because the lessons we learned from it shaped almost every decision we’ve made since, and because I think there’s value in being honest about the times when things go wrong.

The book was a work of narrative nonfiction. I’ll call it “Project Atlas,” though that wasn’t its real title. The author was talented, the subject was timely, and the advance reviews from people who read early copies were enthusiastic. We gave it the biggest launch we’d ever attempted. We printed more copies than we’d printed of any previous title. We spent more on marketing. We booked more events. We committed, financially and emotionally, at a level that was unprecedented for ScrollWorks.

It flopped. Not gently. Not in the way that most books underperform, which is quietly and gradually. Project Atlas flopped loudly and immediately. First-month sales were less than half of what we’d projected. Returns started coming back from bookstores within weeks. By the end of the second month, it was clear that the book was going to lose us a significant amount of money. By the end of the first year, it had lost us more money than any other title in our history, and the financial impact took nearly two years to fully recover from.

What went wrong? I’ve spent a lot of time thinking about this, and I’ve identified at least five distinct failures, each of which taught us something. Some of the failures were ours. Some were the book’s. Some were the market’s. All of them were instructive.

The first failure was timing. Project Atlas dealt with a subject that was in the news when we acquired the manuscript, roughly eighteen months before publication. By the time the book came out, the news cycle had moved on. The subject hadn’t become less important, but it had become less urgent. The sense of “everyone is talking about this” that had made the book feel timely during acquisition had evaporated. This is a risk with any topical nonfiction, and we should have been more aware of it. A book that’s tied to a specific news moment needs to be published quickly, ideally within six months of acquisition. Our eighteen-month timeline was standard for us, but it was too slow for this particular book. By the time we delivered, the window of cultural attention had closed.

The second failure was positioning. We marketed Project Atlas as a book for general readers. The jacket copy, the press materials, the review pitch, all of it was aimed at a broad audience of people who cared about the subject. The problem is that the book itself was more specialized than its marketing suggested. The author had done deep, original research, which was one of the book’s genuine strengths, but the depth of the research also made the book dense and demanding in ways that general readers weren’t expecting. Readers who picked it up based on our marketing found it heavier going than they’d anticipated. Many of them put it down. The people who would have loved the book, specialists and deeply informed general readers with a pre-existing interest in the subject, weren’t reached by our marketing because we’d aimed above them, trying to build a broader audience that the book’s content couldn’t support.

This mismatch between the book and its positioning was our fault, not the author’s. The author wrote the book they wanted to write, and it was a good book. We were the ones who decided it could be marketed to a mass audience. We were wrong. The book needed to be positioned as a serious, in-depth work for informed readers, with marketing that emphasized the quality and originality of the research rather than the timeliness of the subject. If we’d done that, we would have sold fewer copies, but we would have sold them to the right people, and the book’s reputation would have built from there.

The third failure was the cover. I still wince when I think about it. The cover was beautiful. I loved it. The designer loved it. The sales team loved it. The author loved it. Everyone agreed it was one of the best covers we’d ever produced. It was also, in retrospect, completely wrong for the book. The cover suggested a certain kind of reading experience: accessible, contemporary, perhaps slightly journalistic. The actual reading experience was more scholarly, more meditative, more demanding. Readers who were attracted by the cover and the positioning arrived at a book that didn’t match their expectations. The cover was a promise we couldn’t keep.

This taught me something about book covers that I’ve never forgotten. A great cover isn’t just one that looks beautiful or catches the eye. A great cover is one that accurately communicates what the book is. It sets correct expectations. It attracts the readers who will appreciate the book and, just as importantly, it doesn’t attract readers who won’t. A cover that over-promises is worse than a cover that under-promises, because over-promising leads to disappointed readers, negative reviews, and returns. The cover of Project Atlas got people to pick up the book. It didn’t prepare them for what they’d find inside.

The fourth failure was our print run. We printed 8,000 copies, which was more than double our usual first print run. The decision was based on optimistic sales projections that were themselves based on the advance enthusiasm and the (supposed) timeliness of the subject. When the book didn’t sell as expected, we were stuck with thousands of unsold copies. The financial impact was brutal. Each unsold copy represented not just lost revenue but sunk cost: printing, shipping, and warehousing that we’d already paid for. The returns from bookstores compounded the problem, because returns cost money to process and the returned books had to be pulped or remaindered at a fraction of their cost.

We now have a strict rule about first print runs: never print more than 150% of our conservative sales projection, regardless of how optimistic we feel. If a book takes off and we need to reprint, the cost of a rush reprint is less than the cost of warehousing thousands of unsold copies. This rule has saved us from repeating the Project Atlas mistake multiple times. It’s also cost us some sales when books sold faster than expected and we ran out of stock while waiting for a reprint. But the missed-sales cost is always smaller than the overstock cost. Always.

The fifth failure, and this is the one I’m least comfortable admitting, was my own enthusiasm. I loved this book. I thought it was important. I thought the writing was strong, the research was original, and the subject mattered. All of those things were true. But my love for the book clouded my judgment about its commercial potential. I conflated “this book deserves a large audience” with “this book will find a large audience.” Those are very different things. Lots of books deserve large audiences and don’t get them. The market doesn’t reward books based on their merit. It rewards books based on their ability to find and satisfy readers. Project Atlas was a meritorious book that wasn’t accessible enough to satisfy the audience we tried to build for it.

This is the hardest lesson I’ve learned in publishing, and I keep having to relearn it. My personal taste is not a reliable predictor of commercial performance. I can love a book and be right about its quality while being completely wrong about its sales potential. The two judgments require different faculties. Quality assessment is about craft, intelligence, originality, and emotional resonance. Sales prediction is about market conditions, audience size, competition, timing, positioning, and a dozen other variables that have nothing to do with the book’s inherent merit. I’m good at the first. I’m less good at the second. Knowing that about myself has made me a better publisher, because it means I seek out opinions and data that can correct for my biases.

What happened to the author? She was disappointed. Of course she was. She’d worked on the book for four years, and its commercial failure felt personal even though it wasn’t. We had a long, honest conversation about what went wrong. I told her that the positioning and marketing failures were ours, not hers. I told her the book was good, that I still believed in it, and that its commercial performance didn’t define its value. I’m not sure she believed me. Authors hear publishers’ excuses and consolations all the time, and after a while, the words start to sound hollow.

She published her next book with a different publisher. I understood. A publisher that’s lost money on your last book, however good their intentions, carries an association that’s hard to shake. She deserved a fresh start. Her second book did better, though not dramatically so. Publishing is a business where early career stumbles can be overcome. One underperforming book doesn’t define an author, and the readers who did find Project Atlas were passionate about it. The book has a small, dedicated following that occasionally surfaces on Twitter and Goodreads, recommending it to anyone who will listen. In a different timeline, with better positioning and a more accurate cover and a publication date six months earlier, those passionate readers might have been joined by tens of thousands of others.

The experience changed how we make decisions at ScrollWorks. Some of the changes were procedural. The print run caps. The architecture review for positioning. The practice of testing cover designs with readers outside our immediate circle to check whether the expectations set by the cover match the book’s actual content. Other changes were more philosophical. We became more cautious about topical nonfiction, not avoiding it but insisting on a shorter timeline from acquisition to publication. We became more disciplined about separating our enthusiasm for a book’s quality from our assessment of its commercial potential. We started building what I call a “disagreement step” into our acquisitions process, where someone on the team is specifically tasked with arguing against acquiring a book that everyone else loves. The disagreement isn’t meant to kill the acquisition. It’s meant to surface risks that enthusiasm might have hidden.

I’ve also become more comfortable with the idea that not every book needs to reach a large audience. Some books are for 2,000 readers, and that’s fine. Those 2,000 readers might be deeply affected. They might carry the book with them for years. They might recommend it to people who recommend it to other people, building an audience slowly over decades rather than weeks. Literary history is full of books that failed commercially on publication and found their audience much later. Moby-Dick was a commercial disaster. The Great Gatsby sold poorly in Fitzgerald’s lifetime. These are extreme examples, and I’m not comparing Project Atlas to either of them. But the principle applies: a book’s commercial reception at the moment of publication is not the final word on its value.

I still think about Project Atlas regularly. Not with regret, exactly. More with the rueful awareness that we could have done better, and that doing better would have meant being more honest with ourselves about what the book was and who it was for. The best service we could have given that book was accuracy: accurate positioning, an accurate cover, an accurate assessment of its audience size, and an accurate timeline that matched its subject’s position in the cultural conversation. Instead, we gave it ambition. Ambition is a good quality in a publisher, but not when it substitutes for clear thinking.

If you’re a publisher reading this, or a writer, or anyone involved in bringing books into the world, I hope the story is useful. Failures in publishing are common but rarely discussed in public. Success stories are everywhere. Conference talks are full of them. Industry memoirs are built around them. But the failures are where the real learning happens, because failures force you to examine your assumptions and admit that your judgment was wrong. I was wrong about Project Atlas. I was wrong in five specific, identifiable ways. And because I was willing to name those ways and build systems to prevent them, ScrollWorks is a better publisher than it was before. The worst book on our list turned out to be the most valuable one, not because of what it was, but because of what it taught us about who we are and how we work.